The results of long-range financial projections for a Social Security system are inherently uncertain since they depend on complex models that contain estimates of future demographic and economic factors that are projected over a period of years. These factors may vary over the length of the projection period; for example, future mortality rates may be projected to decline as the general health and life expectancy of the population improves over time. In the United States, the official projections for the Social Security system cover a period of seventy-five years and it is widely acknowledged that the degree of reliability of the estimates declines as the length of the projection period is extended from twenty-five to fifty and seventy-five years. The Social Security Trustees present projection results that are prepared by the Office of the Chief Actuary of the Social Security Administration; the Congressional Budget Office prepares its own separate projections using different methodology and assumptions. Deterministic projection methods recognize only a single value for each demographic and economic factor at each point over the projection period, whereas stochastic projection methods recognize a range of plausible outcomes for certain of the demographic and economic factors.
Actuaries, statisticians, demographers and economists all possess special technical skills and training that may be applied to certain aspects of the Social Security financial projections. Actuaries have long been acknowledged as experts in mortality measurement, particularly in connection with the business of life insurance and annuities. Demographers and statisticians possess specialist skills and they utilize advanced techniques to build complex models for population projections and trends in mortality rates. A recent paper Statistical Security for Social Security by two demographers, Samir Sineji and Gary King, has attracted considerable attention and debate for its in-depth analysis of the mortality improvement projection methodology that underlies the financial projections that appear in the Social Security Trustees’ annual reports. The authors of this paper comment on how the current official mortality projections are based on a combination of linear extrapolation and qualitative judgments. The authors explain that linear extrapolation methodology does not adequately recognize known relevant risk factors and produces anomalous results that are inconsistent with well-established demographic patterns, and they further state that modern statistical methods typically outperform even the best qualitative judgments. Their paper advocates the use of these more advanced techniques in the interests of achieving the advantages of transparency, replicability, reduction of uncertainty, and protection of the Social Security system against the possible vulnerability of the official published financial projections to misinterpretation and misrepresentation.
The Office of the Chief Actuary employs a multi-step process to produce age and sex specific mortality forecasts seventy-five years into the future, based on a combination of linear extrapolation of historical data and labor-intensive subjective choices of seventy interrelated ultimate rates of decline. These rates of decline are determined for males and females in each of five defined age groups and for seven different causes of death, namely, heart disease, cancer, vascular disease, violence, respiratory diseases, diabetes mellitus, and a residual category of all other causes. In its analysis of the efficacy of this complex process, the authors of the paper identified patterns across age of ultimate rates of decline that differ considerably from historical patterns for several of the cause-sex groups. Other anomalies that were identified include vacillation in the rate of mortality decline between age groups. The subjective choice of seventy ultimate rates of decline introduces complexity and obscurity into the method. Another criticism of the methodology focuses on independently forecasting age-specific mortality cross-sections, whereas these cross-sections are in fact dependent. Better mortality forecasting approaches are available that formally incorporate additional information while preserving quintessential demographic patterns.
The magnitude of results of financial projections is found to be sensitive to changes in methodology and demographic and economic assumptions. The annual Social Security Trustees’ report discloses the financial effect of changing each of the principal assumptions used in making the financial projections, so that there is some disclosure of the effect of the uncertainty that is inherent in the methodology. While sensitivity is expressed in terms of its effect on solvency duration or the level of financing to restore solvency over a period, the paper by Sineji and King essentially identifies a magnitude of deviation in the official estimates from a more scientific methodology for assessing Statistical Security for Social Security that they estimate to be at $730 billion. The authors provide a number of demonstrations of the validation and assessment of formal statistical mortality forecasts that include direct allowances for certain specific risk factors, such as obesity and smoking. Greater coordination between actuaries, statisticians and demographers would be beneficial in terms of improving transparency and enhancing the reliability and accuracy of Social Security financial projections while reducing the degree of inherent uncertainty.